Share This Author
Explaining Time Series Predictions with Dynamic Masks
With synthetic and real-world data, it is demonstrated that the dynamic underpinning of Dynamask, together with its parsimony, offer a neat improvement in the identification of feature importance over time.
Label-Free Explainability for Unsupervised Models
This work introduces two crucial extensions of post-hoc explanation techniques that respectively highlight inﬂuential features and training examples for a black-box to construct representations at inference time and demonstrates that these extensions can be successfully implemented as simple wrap-pers around many existing feature and example importance methods.
Learning outside the Black-Box: The pursuit of interpretable models
This paper proposes an algorithm that produces a continuous global interpretation of any given continuous black-box function, and demonstrates that the algorithm produces global interpretations that are both highly accurate and parsimonious (involve a small number of terms).
Explaining Latent Representations with a Corpus of Examples
SimplEx is a user-centred method that provides example-based explanations with reference to a freely selected set of examples, called the corpus that improves the user’s understanding of the latent space with post-hoc explanations.
DAUX: a Density-based Approach for Uncertainty eXplanations
- Haoliang Sun, B. V. Breugel, Jonathan Crabbe, Nabeel Seedat, M. Schaar
- Computer ScienceArXiv
- 11 July 2022
This work proposes an assumption-light method for interpreting UQ models themselves and introduces the confusion density matrix—a kernel-based approximation of the misclassiﬁcation density—and uses this to categorize suspicious examples identi ﬁed by a given UQ method into three classes.
Concept Activation Regions: A Generalized Framework For Concept-Based Explanations
This work introduces an extension of the CAV formalism that is based on the kernel trick and support vector classifiers that yields global concept-based explanations and local conceptbased feature importance, and proves that CAR explanations built with radial kernels are invariant under latent space isometries.
Data-SUITE: Data-centric identification of in-distribution incongruous examples
Data-SUITE is empirically validate its performance and coverage guarantees and demonstrated on cross-site medical data, biased data, and data with concept drift, that Data-SUite best identiﬁes ID regions where a downstream model may be reliable (independent of said model).
Benchmarking Heterogeneous Treatment Effect Models through the Lens of Interpretability
This work uses post-hoc feature importance methods to identify features that inﬂuence the model’s predictions and constructs a benchmarking environment to empirically investigate the ability of personalized treatment effect models to identify predictive covariates – covariates that determine differential responses to treatment.